Tor Detection using a Machine Learning Approach Using Correlation based Feature Selection with Best First and Random Forest
Malak Hamad Al-mashagbeh, Mohammad Ababneh · 2021
Since the evolution of the internet and accompanying services and technologies, people all over the world had started enjoying these developments, but unfortunately this wasn't without cost. On one side, companies and service providers had found ways to maximize their benefit from their customers' usage of their services by tracking them and selling their information exposing their privacy. On the other side, criminals had also found their ways to avoid security measures and bypass technologies to conduct their crimes. One of the technologies that was developed to address the previous issues was TOR, which can provide privacy to users and unfortunately, can provide protection for criminals by hiding and disguising their traffic. Technology developers and researchers successfully tried to develop approaches to detect TOR traffic, but these solutions suffered from shortcomings such as low accuracy and efficiency. In this research, we present an ML approach using Correlation-based Feature Selection with Best First and Random Forest. We used a dataset from (UNB CIC) University implemented in WEKA. Our results show that time based characteristics can be used to effectively detect Tor traffic with very high precision and efficiency.